Identity Politics and Trade Policy - GENE M. GROSSMAN - Harvard University
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Copyedited by: ES MANUSCRIPT CATEGORY: Article
Review of Economic Studies (2021) 88, 1101–1126 doi:10.1093/restud/rdaa031
© The Author(s) 2020. Published by Oxford University Press on behalf of The Review of Economic Studies Limited.
Advance access publication 27 July 2020
Identity Politics and Trade
Policy
GENE M. GROSSMAN
Princeton University
Downloaded from https://academic.oup.com/restud/article/88/3/1101/5858143 by guest on 24 May 2021
and
ELHANAN HELPMAN
Harvard University and CIFAR
First version received July 2019; Editorial decision May 2020; Accepted June 2020 (Eds.)
We characterize trade policies that result from political competition when assessments of well-
being include both material and psychosocial components. The material component reflects, as usual,
satisfaction from consumption. Borrowing from social identity theory, we take the psychosocial component
as combining the pride and self-esteem an individual draws from the status of groups with which she
identifies and a dissonance cost she bears from identifying with those that are different from herself. In
this framework, changes in social identification patterns that may result, for example, from increased
income inequality or heightened class or ethnic tensions, lead to pronounced changes in trade policy. We
analyse the nature of these policy changes.
Key words: Social identity, Political economy, Tariff formation, Protectionism, Populism
JEL Codes: F13, D78
1. INTRODUCTION
Gradual trade liberalization proceeded steadily throughout most of the post-war period. But recent
years have witnessed a dramatic reversal of trade policies in some countries. What accounts for
this sudden shift in the political winds?
Many theories of the political economy of trade policy point to the role that special interest
groups play in scoring protection for their members; see, for example, Grossman and Helpman
(2001, 2002). But interest groups do not seem to have played a central part in recent events.
In special interest politics, lobbyists seek preferential treatment for some specific industry or
factor of production at the expense of others. The rhetoric of the recent policy reversals seems
to be addressed to broader segments of the voting population, having as its apparent goal a
repudiation of globalization and an across-the-board reduction in imports. Efforts to understand
the political economy of this about-face might fruitfully focus on factors that led to changes in
(some) voters’ preferences rather than on changes in the behaviour of interest groups. Indeed,
Pavcnik (2017) reports—using data from the Pew Global Attitudes Survey—a decline of about
The editor in charge of this paper was Nicola Gennaioli.
1101
[09:42 8/5/2021 OP-REST200030.tex] RESTUD: The Review of Economic Studies Page: 1101 1101–1126Copyedited by: ES MANUSCRIPT CATEGORY: Article
1102 REVIEW OF ECONOMIC STUDIES
ten percentage points between 2002 and 2014 in the fraction of U.S. respondents that view
globalization favourably, as indicated by a reply of “good” or “very good” to the question, “What
do you think about the growing trade and business ties between the U.S. and other countries, do
you think it is a very good thing, somewhat good, somewhat bad, or a very bad thing for our
country?”
In this article, we explore the idea that voters’ preferences over trade policy reflect not only
their own material self-interests but also concerns for members of those groups in society with
whom they identify. As Shayo (2009) argues, voters often consider the interests of others when
formulating their attitudes about redistributive policies, but such altruism typically is particular
rather than universal. Individuals predominantly care about the well-being of those they perceive
to be similar to themselves. In such a setting, discrete shifts in patterns of social identification
can result in precipitous changes in policy outcomes.
Downloaded from https://academic.oup.com/restud/article/88/3/1101/5858143 by guest on 24 May 2021
Social psychologists define social identity as “the individual’s knowledge that he belongs to
certain social groups together with some emotional and value significance to him of the group
membership” (Tajfel’s 1974, p. 31). Social identity theory builds on the assumption that society
encompasses a variety of social categories that stand in power and status relation to one another
(Hogg and Abrams 1982, p. 14). The pertinent categories are history and context dependent
and, to some extent, fluid (Hogg et al. 1995; Huddy 2001). They may include divisions along
lines of nationality, race, ethnicity, class, occupation, gender, religion, and so on, with different
categorizations acquiring salience in different political, economic, and cultural environments.
A key tenet of the social identity theory developed by Tajfel (1981) and Tajfel and Turner
(1979)—and its close cousin, self-categorization theory as put forth by Turner et al. (1987)—is
that individuals choose the categorical groups with which they associate based on the positive
self-esteem they can derive from such a self-image and their sense of belonging to the group.
Individuals need not be “accepted” into the groups with which they identify, nor need the other
imagined members of an individual’s identity group see themselves similarly. Rather, individuals
compare themselves to a prototypical group member and derive satisfaction from the status that
a group of others with similar social characteristics (along some dimension) enjoys in society.
Akerlof and Kranton (2000, 2010) introduced identity into economics. They posited a utility
function that includes not only the consumption of goods and services but also a psychosocial
component of self-image. Self-esteem, they assumed, is enhanced when an individual conforms
to the behavioural norms prescribed for their self-imagined social groups. Our approach builds on
theirs, and especially on Shayo’s 2009 formulation that incorporates a perceived utility gain when
an individual identifies with a group that enjoys high status, but a loss from identifying with a
group whose prototypical member is very different from that person along relevant dimensions. In
his paper on redistributive taxation, Shayo advances the notion of a “social identity equilibrium,”
in which individuals self-categorize among a set of salient identity groups as a function of the
behaviours and outcomes of others, where behaviour and outcomes are induced by policies and
the policy environment in turn reflects the identity choices made by the individuals together with
their political and economic actions. We will adopt a conceptually similar notion of equilibrium
when thinking about the formation of trade policies, even if our model differs from his in various
details.1
1. Ansolabehere and Puy (2016) develop a spatial voting model with “positional” and “identity” issues in which
parties and voters have exogenous identities and each voter prefers the party that shares her identity, all else the same. See
also Karakas and Mitra (2018), who similarly take patterns of identification among voters and candidates as exogenously
given, while assuming that voters over-reward or under-punish candidates for their policy positions based on an alignment
of cultural identity. Following Shayo (2009), our work is different in allowing self-categorization of identities. Our model
[09:42 8/5/2021 OP-REST200030.tex] RESTUD: The Review of Economic Studies Page: 1102 1101–1126Copyedited by: ES MANUSCRIPT CATEGORY: Article
GROSSMAN & HELPMAN IDENTITY POLITICS AND TRADE POLICY 1103
The hypothesis that self-categorization might affect attitudes toward trade aligns well with
evidence presented by political scientists. For example, Jardina (2019) reports that those who
adopt a “white identity”—a group that disproportionately includes working class men with modest
education—are more likely to espouse protectionist and anti-immigration positions than those
who express less solidarity with their race. Moreover, she finds that these opinions are not primarily
related to economic self-interest. The appeal of protectionism among those who identity in this
way seems more deeply rooted in concerns about the standing of their peer group in local society
and in the global economy. That is, identity predicts attitudes toward trade even after accounting
for a host of other, individual-specific, economic variables.
We conduct our analysis in a simple economic environment that is familiar from the
Heckscher–Ohlin trade model. There are two factors of production in a small country, more-
skilled labour and less-skilled labour, and two goods, an export good and an import-competing
Downloaded from https://academic.oup.com/restud/article/88/3/1101/5858143 by guest on 24 May 2021
good. The country imports the good that uses less-skilled labour relatively intensively at a given
world prices. The more-skilled individuals constitute the upper echelon in society and account for
a minority of the population. They may self-identify as “elite” or “upper-class,” or whatever term
aptly describes a group of highly educated and well-paid individuals. The less-skilled majority
comprises the “working class,” and these individuals too may identify with the fellow members
of their social class, deeming themselves to be, for example, “Main Streeters” or “middle class.”2
In addition, and non-exclusively, individuals of either skill group may choose to identify with
a broader category that comprises the nation as a whole. We will refer to this group as “the
broad nation,” or just “the nation” for short; in the U.S. context, for example, this would mean
seeing oneself as an “American.”3 We follow the social identity theorists in assuming that the
psychological benefit to an individual from identifying with a group is increasing in the perceived
status of the group, which we take to be the material well-being of the prototypical group member.4
The psychological cost of identification is increasing in the difference between an individual’s
own economic and cultural characteristics and that of the average individual in the category,
reflecting the assumption that identifying with others who are different from oneself creates
cognitive dissonance (see Turner et al. 1987; Mcgarty et al. 1992; Hogg 1996; Hogg and Hains
1996).
The polity chooses an ad valorem tariff on the import good. We invoke a political environment
such as the one described by Lindbeck and Weibull (1987), Dixit and Londregan (1996), and
Grossman and Helpman (1996, 2001). In this setting, there are two political parties that differ
in their fixed ideological stances. Voters are heterogeneous in their preferences for the two
parties. The parties use their trade-policy positions to compete for votes. An individual votes
for a party if and only if her preference for its trade platform outweighs her preference for the
ideological position of the rival party. The equilibrium in this political game features convergence
for instrumental policies such as tariffs. Moreover, if the parties view all groups of voters as having
shares an affinity with recent work on the role of “moral values” in politics, in which voters are endowed with either
“individualizing” or “universal” moral values; see, especially, Enke (2019).
2. In the spirit of self-categorization, we should also allow the skilled individuals to self-identify as working class
and the unskilled to self-identify as elite. However, it is reasonable to assume that the dissonance costs of such cross-
identification would be too severe in the light of the underlying differences between the groups to deliver such outcomes
in equilibrium.
3. In political discourse, what it means to be “a real American” may differ across individuals; see, for example,
Huddy (2001, p. 130). For our purposes, we will take identifying with “the nation” to mean identifying with a prototypical
denizen whose characteristics are average in the country’s population.
4. Shayo takes status to be a relative measure that compares the well-being of “in-group” members to that of
the “out-group”; we will discuss this possibility further below and argue that it makes little difference to our qualitative
conclusions.
[09:42 8/5/2021 OP-REST200030.tex] RESTUD: The Review of Economic Studies Page: 1103 1101–1126Copyedited by: ES MANUSCRIPT CATEGORY: Article
1104 REVIEW OF ECONOMIC STUDIES
ideological preferences drawn from a common distribution, then the equilibrium policy is the one
that maximizes a utilitarian social welfare function. With this model in mind, we seek to identify
the tariff that achieves the utilitarian optimum subject to the self-identification constraints. When
we sum voters’ utilities across the polity to form the utilitarian objective, we incorporate not
only the components of welfare that measure individuals’ material well-being, but also those that
measure the psychological satisfaction they derive from their chosen social identifications.5
Our goal is to characterize how trade policy outcomes reflect the economic, political, and
cultural environment. To that end, we define an identification regime to be a complete description
of the pattern of social identification; i.e., a list of all of the groups with which each individual
identifies. In our simple model with two types of individuals distinguished only by their skill levels,
there are a limited number of possible regimes. More-skilled individuals might identify or not as
members of the society’s elite social group and they might identify or not more broadly with the
Downloaded from https://academic.oup.com/restud/article/88/3/1101/5858143 by guest on 24 May 2021
nation as a whole. Similarly, the less-skilled individuals might see themselves as members of the
working class, or as nationals, or neither, or both. Typically, a small change in some parameter
will not induce a change in identification regime. Nonetheless, it may generate a change in
the equilibrium policy, unlike what happens when the psychosocial components of utility are
absent. For example, we find that when every individual identifies with her own social class and
the members of at least one skill group also identify more broadly, then neutral technological
progress that does not alter the identification regime results in an increase in the rate of protection.
This result runs counter to what happens when social identification is absent, in which case the
utilitarian optimum for a small country always manifests free trade.
But we are especially interested in situations when a change in the political or economic
environment alters the identification regime.6 In keeping with recent political history, we focus
specifically on “populism.” In his well-regarded book, Müller (2016) defines populism as a specific
form of identity politics in which a group of voters rejects all claims by the elite members of
society to moral legitimacy. By his description, “populists do not just criticize elites; they also
claim that they and only they represent the true people”(Müller 2016, p. 40). Populists advocate
policies that (in their minds) rightfully neglect the interests of the “corrupt” segments of society
while serving instead the interests of “real” people. Many commentators see the rise in populism
as one of the critical developments in recent European and American political history.
To capture a rise in populism in the context of our model, we posit an initial identification
regime in which less-skilled workers identify not only as “working class,” but also more
inclusively as members of a group that includes all denizens of the country. Such initial conditions
are consistent with the findings reported by Mansfield and Mutz (2009) that Americans’
attitudes toward trade—as reflected in 2004 survey responses—were based as much or more
on respondents’ perceptions of how trade would affect the U.S. economy as a whole as on
their assessments of how trade would affect their own personal incomes. Following a populist
revolution, the identification regime changes. In the event, the less-skilled cease to identify with
the nation, or at least with their former understanding of a broad nation that includes the elites.
Instead, they come to see the nation as synonymous with their own class and type. We describe
the parameter changes in the model that can give rise to such an (endogenous) narrowing of
identification and study the implications for the equilibrium trade policy. Interestingly, we find that
5. As a robustness check, we also consider briefly in the text—and more fully in the Supplementary Appendix—the
tariff policy that maximizes the welfare, economic, and psychological, of the median voter.
6. Atkin et al. (2019) provide evidence that social identities are fungible and respond to changes in the economic
environment.
[09:42 8/5/2021 OP-REST200030.tex] RESTUD: The Review of Economic Studies Page: 1104 1101–1126Copyedited by: ES MANUSCRIPT CATEGORY: Article
GROSSMAN & HELPMAN IDENTITY POLITICS AND TRADE POLICY 1105
if the elite comprise a sufficiently small fraction of the nation’s population, a populist revolution
of this sort induces a discrete rise in the equilibrium tariff rate.7
Our theoretical finding resonates with evidence provided in a recent paper by
Rogowski and Flaherty (2020). They point out that recent anti-globalization movements in Europe
“often excoriate not just globalization or immigration, but allegedly nefarious elites (p. 2).” Using
both cross-country and regional-level voting data for Europe, they find that anti-globalization
parties fare best where income inequality is highest, and that trade shocks have little or no effect
on voting behaviour where inequality is low. If high inequality goes hand in hand with narrow
identification—as our model would predict—then we might indeed expect a stronger backlash
against trade in regions and countries with a wider distribution of income.
Our goal in the article is not to prove that certain changes in identity have been responsible
for recent shifts in trade attitudes and trade policies. Rather, we aim to show that aspects of social
Downloaded from https://academic.oup.com/restud/article/88/3/1101/5858143 by guest on 24 May 2021
identity can readily be incorporated in models of the political economy of trade policy. When this
is done, changes in patterns of social identification can alter preferences about trade policy, both
among particular groups in society and in the aggregate. Moreover, inasmuch as identity politics
builds on a dichotomous distinction between “in-groups” and “out-groups,” the policy response
to changes in social identification can be sudden and dramatic.8
The remainder of the article is organized as follows. In the next section, we introduce social
identity as a component in individuals’ utility that influences their voting behaviour. In Section 3,
we identify the tariff that would emerge from electoral competition in a Lindbeck–Weibull-style
political model and argue that protectionist outcomes can emerge in situations in which free trade
would prevail absent the psychosocial components of individual utility. Section 4 addresses how
small changes in the economic, political, and cultural environment affect trade policy when the
patterns of social identification remain fixed. In Section 5, we consider the policy effects of shifts
in the identification regime and, in particular, those of a populist revolution in which the working
class repudiates its identification with a broad national group that includes the elites. A brief
Section 6 examines the robustness of our results to the choice of political maximand; there we
assume that parties cater to the median voter. Section 7 concludes.
2. A MODEL OF SOCIAL IDENTITY IN TRADE POLITICS
We consider a simple and familiar trading environment. A small country produces and trades
two goods at fixed world prices. The goods are produced competitively, with constant returns
to scale, by two types of individuals: the “more-skilled,” h, and the “less-skilled,” . The less-
skilled workers comprise a majority of the population, so that λ > λh , where λi is the fraction
of individuals of type i. The export good, X, makes relatively intensive use of the more-skilled
labour, while the import-competing good, Z, makes relatively intensive use of the less-skilled
labour. We normalize the size of the population to one.
7. Our explanation for a sharp increase in protectionism resulting from a discrete change in social preferences is
complementary to the explanation offered by Gennaioli and Tabellini (2019). They too focus on shifts in social identity,
but in an environment where identification causes voters to slant their beliefs toward the distinctive opinions of those with
whom they identify. In such a setting, economic shocks that induce changes in the identification regime also generate
adjustments in beliefs about the efficacy of different policies.
8. As noted above, Gennaioli and Tabellini (2019) offer a complementary explanation for dramatic shifts in policy
preferences based on changes in beliefs. See also Besley and Persson (2019), who propose a model of identity politics
in which identification evolves over time and the political parties strategically propose policies to influence the political
salience of different dimensions of identity.
[09:42 8/5/2021 OP-REST200030.tex] RESTUD: The Review of Economic Studies Page: 1105 1101–1126Copyedited by: ES MANUSCRIPT CATEGORY: Article
1106 REVIEW OF ECONOMIC STUDIES
All individuals irrespective of skill level share the same, quasi-linear materialistic utility
function, whereby
νi = cXi +v(cZi ) (1)
is the material utility enjoyed by a representative individual with skill level i who consumes
quantity cXi of the export good and quantity cZi of the import-competing good. The quasi-linear
form allows us to ignore income effects, which are not particularly germane to the analysis, but
are analytically cumbersome. We assume that v(·) is increasing and concave.
An individual’s overall fulfilment comprises the sum of her material utility, as just described,
and a psychosocial component that comes from self-identifying with those groups in society that
have emotional significance to her. Psychosocial utility has two subcomponents, one positive and
one negative. On the one hand, an individual takes pride in seeing herself as a member of a group,
Downloaded from https://academic.oup.com/restud/article/88/3/1101/5858143 by guest on 24 May 2021
the more so the greater is the “status” of that group in its social context. According to Tajfel’s
(1974) social identity theory, identifying with a group of others confers dignity and self-respect.
On the other hand, there is cognitive dissonance that ensues from identifying with others who are
very different from oneself. Self-assurance derives from seeing oneself as fundamentally similar
in relevant ways to those others whom one respects. In contrast, seeing oneself as a member of a
group whose others are very different may undermine an individual’s confidence in her attributes
and behaviour.
We take the benefit to any individual from identifying with some group to be a linearly
increasing function of the status of the average member of that group, where status is measured
by material well-being. The dissonance cost is an increasing, quadratic function of the distance
in some conceptual space between an individual’s own attributes and those of the prototypical
member of a group with whom she identifies. Since the social groups that we will consider in
this article are defined in socioeconomic terms, we measure distance also in units of material
well-being; that is, an individual bears a greater cost from identifying with a group of others who
are, on average, much wealthier or much poorer than she.9 Combining these two components,
we have that a representative individual with skill level i who chooses to self-identify with a
socioeconomic group g gains a net benefit from that identification of
g g g 2 g g
ui = Ai +αi ν̄ g −βi νi − ν̄ g , αi > 0,βi > 0, (2)
g
where Ai is a constant reflecting the baseline level of pride that comes from identifying with
group g for an individual of type i and ν̄ g is the average material utility of all those who are
g
considered by society as having the characteristics that define group g. In principle, Ai could be
negative, if individuals perceive a stigma from identifying with a group of sufficiently low status.
g
The parameter αi can be thought to represent the intensity of identification with group g among
those with skill level i; Huddy (2001, p. 137), for example, has emphasized variation in the degree
to which different types of individuals identify with different salient groups and the importance
of this consideration for political outcomes.10
9. Our use of the Euclidean metric for income differences follows a common practice in the psychology of
identification and categorization; see, for example, Nosofsky (1986) or Ashby (1992). An alternative would be to assume
that dissonance depends on relative well-being, i.e., vi /v̄g . We find little in the experimental literature that would allow us
to distinguish between these alternative hypotheses and adopt the Euclidean measure of distance for illustrative purposes
only.
10. Some experimental evidence suggests the existence of a third, psychosocial subcomponent of utility associated
with identity, namely envy of members of out-groups who enjoy greater material well-being than oneself; see, for example,
Chen and Li (2009) and Kranton et al. (2018). We could readily include such an element by subtracting from ui a term
[09:42 8/5/2021 OP-REST200030.tex] RESTUD: The Review of Economic Studies Page: 1106 1101–1126Copyedited by: ES MANUSCRIPT CATEGORY: Article
GROSSMAN & HELPMAN IDENTITY POLITICS AND TRADE POLICY 1107
Two comments are in order. First, as stressed in the self-categorization theory developed by
Turner et al. (1987), identification is voluntary and unregulated; it is a personal choice made
by the individual and its psychological significance resides only in that person’s own mind. An
individual need not secure anyone’s permission to identify in any particular way, nor can she be
coerced to identify with groups that share her salient characteristics.
Second, the set of groups with whom an individual may identify and the attributes of the
prototypical members of those groups are given at a moment in time by the cultural and historical
context. That is, we take the defining characteristics of group g as exogenous (although they may
change over time in response to political and historical events) when we calculate the average
characteristic in the group, ν̄ g in equation (2); in particular, ν̄ g does not depend on the set of
individuals that choose to identify with group g. While it would be interesting to consider how
the set of social groups that are salient in a society evolves over time, as well as the determinants
Downloaded from https://academic.oup.com/restud/article/88/3/1101/5858143 by guest on 24 May 2021
of a group’s defining characteristics, that is beyond the scope of the present article.11
To summarize, a representative individual in skill group i has preferences that are summarized
by the utility function ui , where
g g g g 2
ui = cXi +v(cZi )+ Ii Ai +αi ν̄ g −βi νi − ν̄ g , (3)
g∈G
g g
Ii = 1 if the individual self-identifies with group g and Ii = 0 if not, and where G is the set of social
groups that is salient in society at that historical juncture. In this article, we take the set of salient
social identity groups to have three elements: (1) a socioeconomic group that we term “the elite”
and designate by ε that has as its archetype one of the higher-earning individuals in society, so that
ν̄ ε = νh , i.e., the material utility of the representative more-skilled individual; (2) a socioeconomic
group that we term “the working class” and designate by ω that has as its archetype one of the
lower-earning individuals in society, so that ν̄ ω = ν , i.e., the material utility of the representative
less-skilled individual; and (3) a social group that we term “the broad nation” and designate by
b that has as its archetype the average citizen in the country, so that ν̄ b = λ ν +λh νh . It follows
that g g g
g 1 if Ai +αi ν̄ g −βi (νi − ν̄ g )2 ≥ 0
Ii = g g g g , i ∈ {h,},g ∈ {ε,ω,b}.
0 if Ai +αi ν̄ −βi (νi − ν̄ g )2 < 0
To avoid a taxonomy that includes unrealistic cases, we introduce parameter restrictions
that rule out certain identification patterns. First, we assume throughout that Aεh > 0 and Aω
> 0;
all individuals take some pride from identifying with others in their own social class. Then,
since νh = ν̄ ε and ν = ν̄ ω , every individual chooses to identify with her own socioeconomic
group, because such identification confers a psychological benefit but imposes no dissonance
cost. Second, we assume that βhω and βε are large enough that no cross-class identification
occurs; i.e., no more-skilled individual identifies as working class and no less-skilled individual
that represents the psychological cost that a lesser-skilled individual bears from jealousy of each high-skilled individual
who earns a higher wage than she, in circumstances where the individual does not identify with the broad nation and thus
considers more-skilled workers to be members of the “out-group.” We will consider this possibility briefly in Section 5
below, where we will argue that envy of out-group members broadens the set of parameters for which a populist revolution
induces a jump in trade protectionism.
11. It is possible that individuals bear a “psychological fixed cost” that depends upon the number of groups with
which they choose to identify, which might limit self-identification on the extensive margin. We do not pursue this idea
here, because we consider only a few large and highly salient identity groups. That is, we do not believe that identifying
with one’s social class impinges upon identifying more broadly with other nationals.
[09:42 8/5/2021 OP-REST200030.tex] RESTUD: The Review of Economic Studies Page: 1107 1101–1126Copyedited by: ES MANUSCRIPT CATEGORY: Article
1108 REVIEW OF ECONOMIC STUDIES
identifies as an elite.12 We are left with two possible outcomes for each skill type: in addition
to self-identifying as elite, the more-skilled individuals might opt to identify with the broad
nation or not (Ibh = 1 or Ibh = 0); and similarly, in addition to self-identifying as working class, the
less-skilled individuals might choose to identify broadly or not (Ib = 1 or Ib = 0). In total, there
are four possible identification regimes, R={(0,0),(0,1),(1,0),(1,1)}, where an identification
regime r ∈ R is an ordered pair in which the first element describes whether or not the more-skilled
individuals identify broadly with the nation and the second element describes whether or not the
less-skilled individuals do so.
We also introduce a symmetry assumption that sharpens some of our results. In particular, we
invoke
Assumption 1 (i) αhε = αω = α and (ii) αhb = αb = α b .
Downloaded from https://academic.oup.com/restud/article/88/3/1101/5858143 by guest on 24 May 2021
With this assumption of symmetric benefits from status, those that are more skilled take the same
pride from a marginal increase in the status of the elites as those that are less skilled take from
a marginal increase in the status of the working class. Moreover, if the two skill groups both
identify as nationals, then those in each group take similar pleasure from a marginal increase in
the status of the nation.13
We turn next to the political environment. As we mentioned in the introduction, we invoke a
political setting such as that described in Lindbeck and Weibull (1987), Dixit and Londregan
(1996), and Grossman and Helpman (1996, 2001).14 In these papers, two political parties
distinguished by their ideological stances compete for votes by announcing their intentions for a
set of pliable policies. The parties adopt policy platforms to maximize their expected vote counts,
anticipating that individuals will cast their votes for whichever party offers a more agreeable
combination of ideological and pliable policies. We assume that the parties are the leaders in this
political game; i.e., they announce their positions in the first stage, before individuals have self-
categorized. In the second stage, individuals make their personal identity choices, anticipating the
economic and psychosocial outcomes that will result if a given policy is implemented. Finally,
each individual votes for the party that offers higher utility, including an ideological component,
a material component, and the two psychosocial subcomponents.15
12. Equivalently, we could assume that Aωh and Aε are zero or even negative, so that the more-skilled individuals
take little or no pride in identifying as working class and the less-skilled individuals take no pride in pretending to be
elites.
13. We allow for the possibility that βhb = βb , because experimental research has found that individuals are more
bothered by “disadvantageous inequality” than by “advantageous inequality”; see, for example, Fehr and Schmidt (1999)
and Charness and Rabin (2002). These findings would suggest that βhb < βb , i.e., that less-skilled workers suffer a greater
psychological cost from having income below the national average than the more-skilled workers suffer from having
income above that average.
14. The mechanisms by which social identity affects trade policy are present for a wide set of political environments
that imply different mappings from voter preferences to policy outcome. We show in Section 7, for example, that the
effects of a populist revolution are qualitatively similar no matter whether the policy outcome maximizes utilitarian
welfare or the welfare of the median voter.
15. We adopt this order of moves in part for the convenience of describing a unique equilibrium. If individuals
commit to their identities before parties make their platform choices, the outcome we describe below continues to be an
equilibrium but additional equilibria are possible whereby the identity choices justify the policy positions, which in turn are
rationally anticipated by the voters. Clearly, the existence of multiple equilibria could also explain discrete jumps in policy,
if “sunspots” cause the outcome to shift from one equilibrium to another. Our assumed ordering of moves makes most
sense if self-identification can change relatively rapidly in response to political rhetoric and economic events. Evidence
of rapid changes in self-identification can be seen in a different context in Hong Kong residents’ reported identification
[09:42 8/5/2021 OP-REST200030.tex] RESTUD: The Review of Economic Studies Page: 1108 1101–1126Copyedited by: ES MANUSCRIPT CATEGORY: Article
GROSSMAN & HELPMAN IDENTITY POLITICS AND TRADE POLICY 1109
To focus on endogenous rates of protection, we take the pliable policy instrument to be an
ad valorem tariff at rate t.16 The small country faces a fixed relative price, q, of imports in terms
of exports. Then the domestic relative price of the import-competing good is the product of the
world price and one plus the tariff rate, or
p = q(1+t). (4)
The political parties vie for votes with their tariff positions. On the one hand, a positive tariff
may appeal to less-skilled workers, because it raises their wage w via the Stolper–Samuelson
channel, it raises the status of the working-class group with whom they identify, and because it
narrows the wage gap and thus the psychic cost to the less-skilled workers of identifying with the
broad nation, if they happen to do so (albeit at the cost of a reduction of the average wage and
Downloaded from https://academic.oup.com/restud/article/88/3/1101/5858143 by guest on 24 May 2021
thus the status benefits of identifying broadly). By offering a higher tariff than its rival, a party
can attract the votes of less-skilled workers that mildly prefer the rival on ideological grounds.
On the other hand, the offer of a higher tariff stands to cost a party support among the highly
skilled, because it reduces the skilled wage wh , reduces the pride the skilled feel from identifying
as elite, and reduces the psychic benefit from identifying broadly, if they happen to do so (while,
meanwhile, reducing the dissonance cost of identifying thusly).
The parties trade-off these considerations, internalizing as well the anticipated response of
identity choices. Within an identification regime, the electoral competition leads to convergence
of positions to the tariff that maximizes the sum of individual utilities. We will find that,
when the parties move first, the (generically) unique political equilibrium occurs at the tariff
that delivers a global maximum of the utilitarian social welfare function across identification
regimes. Note, however, that in this setting of identity politics, the utility level that influences an
individual’s voting behaviour includes not only the material component of well-being, but also
the psychological component associated with self-identification. Therefore, theutilitarian welfare
function of interest here is i λi ui , not the more usual and limited version, λi νi ; compare (6)
and (3) to see the difference between ui and νi .
Let us consider first the tariff that maximizes utilitarian social welfare in a given identification
regime, r. By (4), the tariff determines the domestic relative price p. This price determines in
turn, the domestic output levels, YX (p) and YZ (p), as well as factor prices, wh (p) and w (p), as
in any model with a Heckscher–Ohlin production structure. Specifically, factor prices in units
of the numeraire good X are such that the domestic price in each sector equals the unit cost of
production. Higher prices p generate a higher less-skilled wage w and a lower more-skilled wage
wh , per the Stolper–Samuelson theorem. Given factor prices, which dictate the cost minimizing
production techniques, output levels are such as to clear the two labour markets.
The income of an individual in skill group i includes not only her wage income wi (p), but
also her per-capita share of the rebated tariff revenues,
T (p,q) = (p−q) (p), (5)
as “Hong Konger” or “Chinese”; see The Economist (2019). Moreover, Atkin et al. (2019) show that identification as
Hindu or Moslem in India responded reasonably quickly to price changes associated with the 1991 economic reforms.
16. A long-standing but unresolved issue in the political economy of trade policy concerns the choice of policy
instrument: Why does the government opt for trade policy as the tool to redistribute income? Social identity theory
hints at a possible answer to this puzzle. Perhaps trade policy aligns well with the concerns that define socioeconomic
cleavages, such that opportunistic politicians sometimes can induce a narrowing of identification by the working class
by touting protectionist policies. Other policy instruments might have less clearcut and salient effects on the status of the
two skill groups, so might be less useful for promoting populism.
We note, as well, that if the policy makers have access to an income tax-cum-transfer, our results about the use and
nature of a tariff in political equilibrium will be little changed, provided that the income tax imposes an excess burden.
[09:42 8/5/2021 OP-REST200030.tex] RESTUD: The Review of Economic Studies Page: 1109 1101–1126Copyedited by: ES MANUSCRIPT CATEGORY: Article
1110 REVIEW OF ECONOMIC STUDIES
where (p) ≡ CZ (p)−YZ (p) is the import-demand function and CZ (p) ≡ argmaxcZ v(cZ )−pcZ
is aggregate consumption of good Z. Therefore, the representative individual in skill group i
achieves material welfare
νi (p,q) = wi (p)+T (p,q)+ (p), (6)
where (p) = maxcZ v(cZ )−pcZ is the common per capita surplus from consumption of good
Z. The psychosocial component of utility for an individual in skill group is
2
u (p,q) = Aω
+α ω
ν
(p,q)+Ib
(p) Ab +αb ν̄ b (p,q)−βb ν (p,q)− ν̄ b (p,q) , (7a)
which is the sum of the self-worth she reaps by identifying with other members of the working
class and the net benefit she gains from identifying broadly with all other nationals, if in fact
Downloaded from https://academic.oup.com/restud/article/88/3/1101/5858143 by guest on 24 May 2021
she elects to do so. We write Ib (p) to emphasize the fact that an individual’s decision whether
to identify broadly depends on the domestic relative price, since that price determines her own
income and that of the national average.
Similarly, the psychosocial component of utility for an individual in skill group h is
2
7buh (p,q) = Aεh +αhε νh (p,q)+Ibh (p) Abh +αhb ν̄ b (p,q)−βhb νh (p,q)− ν̄ b (p,q) , (7b)
the sum of the satisfaction she gains from identifying with fellow members of the elite and possibly
from identifying with the nation. Note that (7b) implies that, if a member of the elite identifies
broadly with the nation, she benefits psychologically from an improvement in the material well-
being of the working class, inasmuch as a rise in ν narrows the gap between herself and the
national average. The expression for uh (p,q) does not, however, imply that a members of the
elite would be happy to sacrifice their own well-being in order to improve the lot of the less
fortunate; transfers from the more-skilled to the less-skilled alleviate any cognitive dissonance
from broad identification, but they also reduce the status from being a member of the elite, not to
mention the materialistic component of a skilled-worker’s utility.17
In a competitive economy,
λh wh (p)+λ w (p) = Y (p),
where Y (p) ≡ YX (p)+pYZ (p) is aggregate GDP at domestic prices. Also, average material utility
is a population-weighted average of the material utility of individuals in each skill group; i.e.,
ν̄ b (p,q) = λh wh (p)+λ w (p)+T (p,q)+ (p). Using these observations, and summing across
all voters, we have under Assumption 1 that aggregate utility in identification regime r is
U r (p,q) = λh Aεh +λ Aω
+(1+α) Y (p)+T (p,q)+ (p) (8)
2
+λh Ib,r
h Abh +α b Y (p)+T (p,q)+ (p) −βhb (1−λh )2 δ(p)
2
+λ Ib,r
Ab +α b Y (p)+T (p,q)+ (p) −βb (1−λ )2 δ(p) ,
17. One might question whether individuals actually feel any discomfort from identifying with others who are less
fortunate than themselves. Our assumption is consistent with survey evidence of a sample of American millionaires that
72% of those with assets in excess of $5 million describe themselves as “middle class” or “upper middle class”; see Frank
(2015). This finding points to discomfort from seeing oneself as different from others, even if that difference is positive.
Moreover, our assumption is consistent with experimental evidence provided by Chen and Li (2009). They reported that
individuals exhibit significantly more charity to those with lower payoffs than themselves when those others are members
of their same identity group.
[09:42 8/5/2021 OP-REST200030.tex] RESTUD: The Review of Economic Studies Page: 1110 1101–1126Copyedited by: ES MANUSCRIPT CATEGORY: Article
GROSSMAN & HELPMAN IDENTITY POLITICS AND TRADE POLICY 1111
where δ(p) ≡ wh (p)−w (p) > 0 is the earnings gap between more-skilled and less-skilled workers
and the Ib,r
i in (8) indicates the broad identification or narrow identification by skill group i in
regime r. The first line on the right-hand side of (8) is aggregate welfare from all material utilities
and from individuals’ identification with their own social class. The following two lines give the
aggregate psychological gain to the more-skilled and less-skilled individuals, respectively, from
identifying as nationals, if regime r is such that members of those skill groups choose to do so.
We offer two observations about the political maximand, U r (p,q). First, when all individuals
identify with their own social class, this introduces an additional element of class warfare to
the political struggle over trade policies. Not only does a less-skilled worker favour an import
tariff to boost her own wage, but she also favours protection to benefit others like herself.
Similarly, a more-skilled individual opposes protection not only to preserve her own pay, but
also to safeguard the incomes of other elites with whom she identifies. However, under the
Downloaded from https://academic.oup.com/restud/article/88/3/1101/5858143 by guest on 24 May 2021
symmetry assumption, these selfish motives for trade policy just offset one another in the calculus
of utilitarian maximization. The selfish aims of the less-skilled individuals who identify with
the working class add λ α w (p)+T (p,q)+ (p) to the political objective, whereas those of
the more-skilled individuals who identify with the elite add λh α wh (p)+T (p,q)+ (p) to the
objective. Together these sum to α Y (p)+T (p,q)+ (p) , which is proportional to aggregate
material welfare. So, this extra component of policy preferences does not tilt the trade politics in
one direction or the other.18
Second, when either skill group identifies broadly with the nation as a whole, this introduces
a source of inequality aversion into the trade politics. Here, such a distaste for inequality does
not reflect a sense of fairness or altruism on the part of any voter. Rather, inasmuch as everyone
pays a psychological toll from identifying with others that are different from themselves, those
that identify with the nation selfishly lean to policies that narrow the gap between themselves
and the average, all else the same. In the present context, a tariff has such an effect, because
the Stolper–Samuelson theorem ties the wage gap to the price of the import-competing good. Of
course, the extent of the “leanings” may differ between those with above-average versus below-
average incomes (as they will if βb = βhb ) and the elites are likely to continue to oppose protection
once all components of their utility are taken into account.
3. EQUILIBRIUM PROTECTION WITH SOCIAL IDENTIFICATION
In this section, we characterize the equilibrium tariff. We describe the conditions under which the
trade politics with social identification give rise to protectionist outcomes, In Section 4, we will
discuss how the equilibrium tariff rate responds to changes in social attitudes, technology, and
the terms of trade for a given identification regime. Then, in Section 5, we will consider changes
in patterns of identification.
Consider Figure 1. The figure depicts U r (p,q) as a function of p for each of the four possible
identification regimes, r. The curve labelled r0 corresponds to the identification regime r = (0,0)
in which no voter identifies broadly with the nation; i.e., Ibh = 0 and Ib = 0. The curves labelled
r and rh correspond, respectively, to identification regimes in which only the lesser-skilled
individuals and only the higher-skilled individuals identify broadly with their fellow nationals,
while those in the remaining skill group identify narrowly only with others in their own social
class. Finally, the curve labelled rh, corresponds to the regime r = (1,1) in which all citizens
18. In contrast, when the political competition maximizes the welfare of the median voter, identification with own
social class favours the working class, because the less-skilled individuals are more numerous and so the median voter is
one of them; see Section 6.
[09:42 8/5/2021 OP-REST200030.tex] RESTUD: The Review of Economic Studies Page: 1111 1101–1126Copyedited by: ES MANUSCRIPT CATEGORY: Article
1112 REVIEW OF ECONOMIC STUDIES
Downloaded from https://academic.oup.com/restud/article/88/3/1101/5858143 by guest on 24 May 2021
Figure 1
Equilibrium protection
identify not only with their own social class, but also with the nation as a whole. We have drawn
the figure under the assumption that all four of these curves are single-peaked as a function of p.
We maintain this assumption henceforth.
We will argue now that the equilibrium tariff is the one that generates the domestic price p that
achieves the global maximum in Figure 1. This is so, because the parties anticipate the identity
choices induced by different policy options. As we shall see, when a party chooses the tariff that
attracts the largest number of voters, it need not worry separately about the individual-rationality
constraints associated with identity choice, because these constraints are always satisfied at p◦ ≡
argmaxp U (p,q), where U (p,q) ≡ maxr U r (p,q). The figure depicts a global maximum at ph, ,
but maxima at ph , p , or p0 are also possible, depending upon the perceived benefits from broad
identification that are reflected in the parameters Abh and Ab , as well as the other parameters that
characterize the economic and cultural environment.
To see that the self-identification constraints always are satisfied at p◦ , suppose first that the
global maximum is achieved for p = ph, , as drawn in Figure 1, where ph, is the domestic price that
maximizes U (p,q) for r = (1,1). Then, U
r (1,0) ph, ,q < U (1,1) ph, ,q and U (0,1) ph, ,q <
U (1,1) ph, ,q ; i.e., the curves rh and r pass below rh, at p = ph, . But the difference between
U (1,1) (p,q) and U (1,0) (p,q) at a given price p is the net (positive) psychosocial benefit to more-
skilled workers from identifying broadly with the nation. Since rh, is above rh at ph, , the
more-skilled will choose to identify broadly at this price. Similarly, rh, above r at ph, implies
that the less-skilled workers also prefer to see themselves as part of the broad nation. In short,
when the parties set the trade policy to achieve a utilitarian maximum at ph, , the identity choices
that underlie these positions are fulfilled.
Now suppose instead that the global maximum is achieved at p = ph , which is the price that
generates the peak of the curve rh . In such circumstances, the fact that U (0,0) (ph ,q) < U (1,0) (ph ,q)
implies that the more-skilled workers enjoy a net benefit from identifying broadly, whereas the
fact that U (1,1) (ph ,q) < U (0,1) (ph ,q) implies that the less-skilled workers would suffer a net loss
from doing so. Again, the welfare-maximizing tariff induces an identification pattern that justifies
the policy choice. An analogous argument applies to the case in which the global maximum occurs
at p = p .
Finally, suppose that the global maximum is attained for p = p0 , which achieves the peak
of the curve r0 . Then U (1,0) (p0 ,q) < U (0,0) (p0 ,q) and U (0,1) (p0 ,q) < U (0,0) (p0 ,q). No workers
would choose to identify broadly under such conditions. In short, no matter which identification
[09:42 8/5/2021 OP-REST200030.tex] RESTUD: The Review of Economic Studies Page: 1112 1101–1126Copyedited by: ES MANUSCRIPT CATEGORY: Article
GROSSMAN & HELPMAN IDENTITY POLITICS AND TRADE POLICY 1113
regime r = r ◦ gives rise to the global maximum of U (p,q), the associated domestic price p◦ =
argmaxp U (p,q) induces self-categorization that justifies the parties’ vote-maximizing policy
choices.
We examine now the nature of the trade policy that underlies each of the possible choices of
the domestic price, p◦ . Suppose first that the global utilitarian maximum is achieved at p0 , so
that each worker identifies narrowly only with her own social class. In this case, U (0,0) (p,q) =
λh Aεh +λ Aω +(1+α) Y (p)+T (p,q)+ (p) , and the equilibrium trade policy is the same as
the one that maximizes aggregate material welfare, Y (p)+T (p,q)+ (p). For a small country,
the implied equilibrium tariff is t ◦ = 0; i.e., the political equilibrium involves free trade. Whereas
the less-skilled workers favour tariffs (up to a point), both for their own real-income gains and
in order to advance the status of the working class with whom they identify, the more-skilled
workers take the opposite position. Indeed, all voters feel more strongly about the issue than they
Downloaded from https://academic.oup.com/restud/article/88/3/1101/5858143 by guest on 24 May 2021
would in the absence of identity concerns. Still, the balancing of these opposing forces based on
self-interest and concern for the status of one’s own social group results in a neutral policy with
respect to trade.
Now suppose that the highest peak occurs for p = p , at which the less-skilled workers identify
broadly with the nation, while the more-skilled workers do not; i.e., r = (0,1). Then the political
maximand comprises the first and third rows in (8), with Ib,r = 1. In addition to constants,
the maximand includes a component that is proportional to aggregate material welfare and a
component that is decreasing in the wage gap, δ(p). The latter reflects the psychological cost to
the less-skilled workers from identifying with other nationals who are different from themselves
in terms of material well-being. As we noted before, the dissonance costs induce an aversion to
inequality on the part of the working class, who suffer from seeing themselves as very different
from the average in a cohort with which they identify. Since a marginal change in the tariff has a
negligible effect on aggregate material welfare at t = 0, and since a tariff reduces the wage gap via
the Stolper–Samuelson mechanism, the maximization of U (0,1) (p,q) generates a positive tariff
in this case, as long as βb > 0.
To understand the source of the protectionist outcome, let us compare the policy preferences
under r = (0,1) to those for the case of r = (0,0) in which neither group identifies broadly. In both
situations, the less-skilled workers see a personal, material benefit from a positive tariff whereas
the more-skilled workers frown upon such a policy for personal, materialistic reasons. As we
have noted, these offsetting preferences just balance in the political calculus of the competing
parties. Similarly, a tariff enhances the self-esteem of the less-skilled workers that derives from
their identifying with the working class, whereas a tariff harms the psyches of the more-skilled
workers, who suffer from the reduced status of elites. Again, these effects offset one another
with respect to the parties’ electoral incentives to support protection. What distinguishes the two
regimes is the pride that the less-skilled enjoy from identifying broadly and the cost they bear
from doing so, which affects policy preferences in r = (0,1) but not in r = (0,0). In r = (0,1), a
small positive tariff generates a second-order loss in the average wage, but a first-order reduction
in dissonance cost. It follows that a (small) positive tariff will attract more votes from the less
skilled when they identify broadly than when they do not. This tilts the political calculus in favour
of protection.
Surprisingly, perhaps, the political equilibrium involves positive protection also in the
identification regime in which only the more-skilled workers identify broadly. With r = (1,0),
the political maximand comprises the first two rows in (8), with Ib,r h = 1. Again, the maximand
combines a term that is proportional to aggregate material welfare and a term that is decreasing
in the wage gap. The former is maximized by free trade, but a small positive tariff shrinks the
[09:42 8/5/2021 OP-REST200030.tex] RESTUD: The Review of Economic Studies Page: 1113 1101–1126Copyedited by: ES MANUSCRIPT CATEGORY: Article
1114 REVIEW OF ECONOMIC STUDIES
dissonance cost for the elites while generating only a second-order loss in aggregate welfare. It
follows that the competing parties will be pushed to a positive tariff in this case as well.
The dissonance costs borne by the elites induce an aversion to inequality when these workers
identify broadly with their fellow nationals. This aversion to inequality does not reflect any sense
of “fairness” or generosity on the part of these elites, but a selfish desire to be “more like the
others.” To be clear, broad identification by the high-skilled workers does not make them favour
protection; their personal, material interests, and their pride of membership in an elite social
class makes them more inclined to vote for the party that proposes the smaller tariff (or bigger
import subsidy). However, the alleviation of dissonance costs generated by a tariff makes each of
them a little less opposed to protection than when they identify narrowly, and so any given tariff
loses fewer votes among the swing voters in this group. Inasmuch as the more-skilled workers
opposition to protection just balances the less-skilled workers support for tariffs when all groups
Downloaded from https://academic.oup.com/restud/article/88/3/1101/5858143 by guest on 24 May 2021
identify narrowly, the outcome tips in favour of protection when the skilled workers identify with
the nation and so become slightly less opposed.
Finally, if broad identification is widespread among both skill groups (i.e. r = (1,1)), the
aggregate aversion to inequality is all the greater. Members of both classes can mitigate their
dissonance cost by narrowing the income gap between themselves and the average. This does
not mean that both groups favour tariffs; in fact, the more-skilled individuals would prefer to see
the tariff reduced from its equilibrium level, to further their own material interests and to boost
the status of the other elites with whom they identify. But the material interests of the more-
skilled voters balance those of the low-skilled voters at t = 0 as do the psychosocial interests from
identifying with own social class, whereas the aggregate marginal effect of broad identification
by the two skill groups tilts toward protection. We recap these findings in19
Proposition 1 Suppose that βhb > 0 and βb > 0. If neither skill group identifies broadly with the
nation, the equilibrium tariff is zero. Otherwise, it is positive.
4. COMPARATIVE STATICS IN A FIXED IDENTIFICATION REGIME
We ask next how small changes in the economic and political environment affect the equilibrium
policy when the identification pattern does not change. That is, we begin from an equilibrium
in which the equilibrium policy is some t ◦ and the identification regime is some r ◦ ∈ R. We
then change one of the parameters of the model by a small amount, such that the equilibrium
identification regime continues to be r ◦ , and examine the response of t ◦ . We consider in turn
g
changes in the psychological cost of identification, βi , changes in the production technologies
(as described further below), and changes in the terms of trade, q. In each case, we employ the
usual method of comparative statics: we calculate the shift in the marginal benefit from a tariff,
evaluated at t = t ◦ , and rely on the second-order condition for an optimum to tell us in which
direction the equilibrium tariff must adjust. From (4), we have that the marginal political benefit
19. Technically, we have made only a local argument that U r (p,q) is increasing in p at p = q. But, it is easy to see
from (9) below that Upr (p,q) > 0 for all p < q. So a global maximum of U r (p,q) cannot be achieved with any t < 0. We
know that a global maximum exists, because for p large enough, either the economy remains incompletely specialized and
δ(p) = 0 or the economy specializes in producing good Z and δ(p) > 0. If the economy remains incompletely specialized,
the optimal domestic price must be smaller than the one that delivers δ(p) = 0. If the economy becomes completely
specialized for large p, wages are proportional to p and so δ (p) is a positive constant. Then the optimal p is below the
lowest price that leads to complete specialization.
[09:42 8/5/2021 OP-REST200030.tex] RESTUD: The Review of Economic Studies Page: 1114 1101–1126You can also read
